Multi-axis load sensors are advantageously used in a variety of application fields. With reference to some of the accompanying drawings, description will be made on a multi-axis load sensor applied by way of example for the force feedback control of an assembly robot.
FIG. 1 is a side view showing an assembly operation by the assembly robot. In the drawing, numeral 1 indicates a hand of the assembly robot while numeral 2 indicates a grip provided with the free end of the hand 1. The drawing illustrates an operation in which a pin 3 is held by the grip 2 and the pin 3 is then inserted in a hole 5 of a counterpart 4 of the fitting. In order to carry out such an operation, it is possible to chamfer the pin 3 and the edge of the hole 5 to suitable extents in advance so that, after moving the pin 3 to a substantially correct position by driving the hand 1 under control, the pin 3 may be inserted into the hole 5 owing to its own weight only by inserting the pin 3 partially into the hole 5 as illustrated in FIG. 1. It is however impossible to insert the pin 3 into the hole 5 by moving the pin 3 to an approximate position and then inserting a part of the pin 3 into the hole 5 in the same manner as mentioned above, where the dimensional difference between the inner diameter of the hole 5 and the outer diameter of the pin 3, namely, the fit clearance is of the level of 1/1000-1/10000 of the inner diameter of the hole 5. Therefore, it becomes indispensable to achieve a precise control with respect to the positioning of the pin 3. It is however impossible, in the light of the current technical level, to control the positioning operation so precisely that the pin can be successfully inserted into its corresponding hole even if the fit clearance is of the above-described minute level. With a view toward overcoming this problem, researches have been carried out on the control which makes use of a multi-axis load sensor 6 provided at a location on the hand 1 as illustrated in FIG. 1. In other words, it has been attempted to control the hand 1 in such a way that the magnitude, direction and point Q of action of a force f produced in accordance with the manner of contact between the pin 3 and hole 5 are determined on the basis of signals detected by the multi-axis force sensor 6 and the manner of action of the load f is adjusted suitably to achieve an ideal state of fitting in accordance with the information on the magnitude, direction and point of action of the force f. In general fitting works, loads applied to work items are unidirectional forces in many instances.
The control of the hand 1 which control makes use of a multi-axis load sensor 6 may be applied not only in such a pin-inserting operation but also in a teaching operation for a robot. The teaching operation for a robot has generally be carried out by an expert, who inputs data, instructions, etc. in accordance with the details of work by operating keys on a teaching panel. However, such a teaching operation requires an expert and takes many hours. In the case of a robot equipped with a hand 1 which is in turn provided with a multi-axis load sensor 6, the teaching can be carried out without operating the keys on the teaching panel provided that a worker moves the hand 1 in accordance with an actual work routine by directly holding the hand 1, the robot is controlled by signals detected by the multi-axis load sensor 6 in the course of the above-described worker's teaching operation, and the control signals are stored. It is also worth mentioning that the above teaching operation can be successfully carried out in a short period of time by an ordinary worker who does not have any special knowledge.
This multi-axis load sensor 6, which is utilized in such fields as described above, is required to have a function capable of detecting at least the magnitude, direction and line of action (which will be described later) of an applied force. Therefore, a discussion will next be made about matter required to detect these magnitude, direction and line of action of the force, based on the vectors of the force f shown in FIG. 2. Considering first the existence of coordinate axes x, y and z which are perpendicular to one another and determine, at their origin P, the magnitude, direction and point Q of action of a force f applied to the point Q of action. It is also to be assumed that the origin P and the point Q are connected to each other by a rigid member. Supposing that the components of the force f in the x-axis, y-axis and z-axis by f.sub.x, f.sub.y and f.sub.z respectively, the magnitude and direction of the force f can be obtained by synthesizing its components f.sub.x, f.sub.y and f.sub.z. Therefore, the magnitude and direction of the force f may be determined if each of the components in the above directions can be detected. by expressing detection values of the components in the above directions as F.sub.x, F.sub.y and F.sub.z respectively then EQU F.sub.x =f.sub.x, F.sub.y =f.sub.y, and F.sub.z =f.sub.z (i)
It is possible to determine the magnitude and direction of the force f if those F.sub.x, F.sub.y and F.sub.z (hereinafter called "F.sub.i " as a whole) are detected.
Next, the position of the point Q of action is to be determined. Supposing now that the distances from the origin P to the point Q of action in the directions of the x-axis, y-axis and z-axis be respectively l.sub.x, l.sub.y and l.sub.z and the moment components caused by a force f along the x-axis, y-axis and z-axis be M.sub.x, M.sub.y and M.sub.z, the following relationship may be established among the distances l.sub.x, l.sub.y and l.sub.z, the moment components M.sub.x, M.sub.y and M.sub.z (hereinafter called "M.sub.i " as a whole) and the aforementioned force components F.sub.x, F.sub.y and F.sub.z : EQU M.sub.x =-F.sub.y .multidot.l.sub.z +F.sub.z .multidot.l.sub.y (ii) EQU M.sub.y =-F.sub.z .multidot.l.sub.x +F.sub.x .multidot.l.sub.z (iii) EQU M.sub.z =-F.sub.x .multidot.l.sub.y +F.sub.y .multidot.l.sub.x (iv)
It is impossible to derive the position (l.sub.x, l.sub.y, l.sub.z) of the point Q of action only from the force components F.sub.i and moment components M.sub.i. It can be known from the above equations only that the position (l.sub.x, l.sub.y, l.sub.z) is located on a certain specific line (which will hereinafter be called "the line of action of a force"). In order to find out the actual point Q of action, it is necessary to know the position, shape and spatial orientation of a body to which the force is being applied and to determine the crossing point between the body and the line of action of the force. This is now explained with reference to a simple example illustrated in FIG. 3, in which numeral 6 is a multi-axis load sensor and letter G indicates a rod-like rigid body fixed on the multi-axis load sensor 6. The rod-like rigid body G extends on and along the z-axis. Letter H indicates a ball having a radius r and attached fixedly to the upper extremity of the rod-like rigid body G. The distance between the center of the multi-axis sensor 6 and the center of the ball H is z.sub.o. Supposing now that a load f is being applied to the ball H, the position (l.sub.x, l.sub.y, l.sub.z) of the point Q of action which point Q is believed to be on the spherical surface of the ball H satifises the following equation: EQU l.sub.x.sup.2 +l.sub.y.sup.2 +(l.sub.z -z.sub.o).sup.2 =r.sup.2 (v)
By solving the equations (i)-(v), the position (l.sub.x, l.sub.y, l.sub.z) can be determined. Once the position, shape and spatial orientation of a body to which the force is being applied (the ball H in FIG. 3) are determined, the position of the point Q of action can be obtained by detecting the force components F.sub.i and moment components M.sub.i by means of the multi-axis load sensor 6. In this case, the position of the point Q of action can still be obtained even if either one of the above six equations (i)-(iv) is not available. When one relies upon the equations (i)-(iv) only, it is possible to know that the force f lies on a line L but it is impossible to specify the point Q of action. However, the position of the point Q of action may generally be determined so long as the line of action of the force can be determined, in other words, the force components F.sub.i and moment components M.sub.i can be detected, because information on the orientation, shape and spatial position of a body to which the force is being applied, for example, the hand of a robot, a body held by the hand or the like can be readily obtained.
Such a multi-axis load sensor as shown in FIG. 4 has conventionally been proposed to detect these force components F.sub.i and moment components M.sub.i. The outline of the construction of the multi-axis load sensor will hereinafter be described.
FIG. 4 is a perspective view illustrating one example of conventional multi-axis load sensors. In the drawing, numeral 7 indicates a first ring connected to the first rigid body (not illustrated). On the other hand, designated at numeral 8 is a second ring which is connected to a second rigid body (not shown) and provided in face-to-face relationship with the first ring 7. Numeral 9 indicates flexible beams which connect the first ring 7 and second ring 8 to each other. Three flexible beams 9 are provided in total. Designated at numeral 10 is a tensile/compression force detection gauge applied on the inner surface of each flexible beam 9. Numeral 11 indicates a shear force detection gauge applied on the outer surface of each flexible beam 9.
In the above construction, it is now presumed that a certain load has been applied, for example, to the first rigid body. The load is transmitted via the first ring 7, each flexible beam 9 and the second ring 8 to the second rigid body. In the course of transmission of the above load, each of the flexible beams 9 is deformed in proportion to a load applied thereto. This deformation of each flexible beam 9 is detected by its respective detection gauges 10,11. Namely, the tensile/compression force detection gauges 10 detect principally the force components F.sub.x,F.sub.y and the moment component M.sub.z while the shear force detection gauges 11 detect mainly the force component F.sub.z and the moment components M.sub.x,M.sub.y. Detection of each of the components F.sub.i and M.sub.i is carried out by inputting a signal from each of the detection gauges 10,11 of each flexible beam 9 to a computer and then performing a prescribed operation. The components F.sub.i and M.sub.i can thus be detected in the above manner, whereby permitting to know the magnitude, direction and line of action of the force applied to the first rigid body. Furthermore, the position of point of action of the force can also be determined if the orientation, shape and spatial position of the first rigid body are known.
However, such a conventional multi-axis load sensor is accompanied by such drawbacks that will be described below.
(1) Each flexible beam undergoes a deformation by at least two components of the force components F.sub.i and moment components M.sub.i of a load when the flexible beam is applied with the load (interaction). Accordingly, signals produced by each detection gauge contain a plurality of components. It is thus necessary to perform a complex operation by means of a computer, if one wants to obtain individual components from the plurality of components. Namely, the above interaction is not a linear interaction in many instances. The operation per se becomes extremely complex unavoidably, if one tries to enhance the accuracy of the operation so as to separate signals. Thus, the time required for the above operation is unignorably long even if a computer is relied upon. When such a multi-axis load sensor is applied to a robot for assembly work or the like, the multi-axis load sensor develops a fatal shortcoming that it is difficult to perform the real time control of the robot.
(2) In order to improve the sensitivity of the multi-axis load sensor, the rigidity of each flexible beam must be reduced. If the rigidity is lowered, the strength of the multi-axis load sensor which is interposed in the transmission path of a load from one of the rigid bodies to the other rigid body is lowered, thereby rendering the multi-axis load sensor unsuitable for actual use. In order to enhance the strength of the multi-axis load sensor on the other hand, it is necessary to make the rigidity of each flexible beam greater. However, an improvement to the rigidity of each flexible beam leads without exception to a reduction in the sensitivity of the multi-axis load sensor. This situation will be explained further, supposing that a multi-axis load sensor is provided with the hand of a robot. When the rigidity of each flexible beam is reduced, the hand of the robot is by itself lowered in strength. If the rigidity of each flexible beam is increased as opposed to the above situation, the detection accuracy is lowered, thereby making the intended control of the robot difficult.
(3) Even if the rigidity of each flexible beam is lowered and the sensitivity of the multi-axis load sensor is hence improved without paying attention to the imminent strength reduction of the multi-axis load sensor, the multi-axis load sensor develops a phenomenon that it is deformed in a direction different from the direction of a load due to the reduced rigidity of each flexible beam, in other words, the multi-axis load sensor shows poor "persistence" characteristics. A multi-axis load sensor having such characteristics affects inversely on the magnitude and/or direction of each load per se. When such a multi-axis load sensor is used to control a robot, it is impossible to perform its control with high accuracy.
There has also been proposed another type of multi-axis load sensor. It has such a structure that a first and second annular members constructed in much the same way as depicted in FIG. 4, namely, a first annular member formed of two rings connected together by a plurality of flexible beams and a second annular member formed of two rings having inner diameters greater than the outer diameter of the annular member and connected together by a plurality of flexible beams are provided, the first annular member is received in the second annular member, the first and second annular members are connected respectively to the first and second rigid members, and the first and second annular members are connected together in such a way that the transmission of each load between the first and second rigid members is effected by way of the first and second annular members. However, a multi-axis load sensor having such a structure as recited above is not essentially different from the multi-axis sensor shown in FIG. 4 and is thus accompanied by similar drawbacks.
As a further example of conventional multi-axis load sensors, reference may for example be made to a multi-axis load sensor having such a structure as disclosed in Japanese Patent Laid-open No. 39079/1976. Namely, the multi-axis load sensor has the following structure. A hub is enclosed in a housing. A spider is provided with one end (upper end) of the hub. A plurality of arms extend out from the spider. The thus-extended arms are each connected to its corresponding detector which is provided fixedly on the housing, whereby detecting each deformation of the arm. On the other hand, a radially-extending flange member is attached to the other end (i.e. the lower end) of the hub. The flange member are connected to detectors which are also fixedly provided on the housing. Accordingly, it is possible to detect displacements of a given pair of mutually-opposing points on the flange member. The housing and hub are connected via a cantilevered bar fixed to the hub. This connection between the housing and the hub is established in such a way that they are allowed to move freely in directions parallel to the longitudinal axis of the cantilevered bar but the cantilevered bar resists against any displacements of the housing and hub in directions perpendicular to the longitudinal axis of the cantilevered bar.
Such a structure is not different in nature from the multi-axis load sensor illustrated in FIG. 4. Due to inclusion of connecting and/or attachment parts between the housing and hub, the hub and spiders, and the hub and flange member, those connecting and/or attachment parts add undesirable characteristics such as non-linear characteristic and hysteresis characteristics to the detection characterstics of the multi-axis load sensor. Where these undesirable characteristics appear to considerable extents, their influence cannot be completely wiped out even if any special operation is performed, whereby reducing the accuracy of the multi-axis load sensor to a significant extent. If such a multi-axis load sensor is applied to a robot, the controllability of the robot is considerably lowered.